Application of Isokinetic Dynamometry Data in Predicting Gait Deviation Index Using Machine Learning in Stroke...

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Data Science

Background:

• Three-dimensional gait analysis is vital for post-stroke hemiplegic patient rehabilitation.
• Sensor data complexity hinders clinical interpretation of gait analysis.
• The Gait Deviation Index (GDI) simplifies pathological gait quantification.

Purpose of the Study:

• To predict the GDI in hemiplegic patients using sensor-acquired isokinetic muscle strength data and machine learning.
• To integrate Biodex dynamometry and Vicon motion capture system data for enhanced gait analysis.

Main Methods:

• A cross-sectional study of 150 post-stroke hemiplegic patients.
• Collected isokinetic muscle strength data (peak torque, work, power, etc.) at various angular velocities.
• Employed Lasso Regression, Random Forest (RF), Support Vector Regression (SVR), and BP Neural Network models for prediction and validated using MSE, R², and MAE.
• Utilized SHapley Additive exPlanations (SHAP) for model interpretability.

Main Results:

• The Random Forest (RF) model demonstrated superior GDI prediction performance (R²=0.89, MSE=16.18, MAE=2.99).
• Key predictors identified by SHAP analysis included maximum work of extensor muscles at 60°/s and 120°/s.
• Muscle strength metrics at varying speeds are crucial for gait rehabilitation assessment.

Conclusions:

• Integrating sensor technology and machine learning provides a powerful tool for analyzing complex clinical gait data.
• A novel, streamlined model for GDI prediction using isokinetic dynamometry data aids rehabilitation progress assessment in stroke patients.
• This approach holds significant potential for broader clinical applications in rehabilitation.